Skip to playerSkip to main contentSkip to footer
  • 5/15/2025
Mistral unleashes a powerful new AI that outperforms GPT-4o and Claude 3.7 β€” all while costing less than DeepSeek! Discover how this game-changing AI is revolutionizing the industry and setting new standards in performance and affordability. Don’t miss the latest in AI innovation! πŸš€βš‘

#MistralAI #GPT4o #Claude3 #AIRevolution #ArtificialIntelligence #TechNews #Innovation #MachineLearning #AI #DeepSeek #FutureTech #AIbreakthrough #NextGenAI #TechUpdate #CuttingEdge #AIcommunity
Transcript
00:00Mistral just launched Medium 3, a frontier-class AI model that outperforms GPT-4.0 and Claude 3.7
00:10Sonnet in coding, languages, and even multimodal tasks while costing a fraction to run. It hits
00:16over 90% of Claude's performance for just $0.40 per million input tokens and runs smoothly on
00:23only four GPUs. And now that it powers LayChat Enterprise with deep integrations, privacy-first
00:29architecture, and no-code AI agents, OpenAI finally has a serious competitor coming straight out of
00:36Europe. The catchphrase Mistral's own research blog led with is, medium is the new large, and the
00:42company is leaning into that pretty hard. Medium 3 sits between their featherweight small and whatever
00:49large. Surprise, they're teasing for later. But don't let the name fool you. Internally, it's what
00:55they call a frontier-class model. It delivers performance that lands in the same neighborhood
00:59as Anthropik's Claude 3.7 Sonnet, Cohere's Command-A, Meta's Llama 4 Maverick, and even OpenAI's
01:07freshly announced GPT-4.0 while pulling that off on a much skinnier compute diet. In plain English,
01:13you can wedge this thing into a four GPU on-prem rig or spin it up in a cloud VPC and still crank out
01:21results that would usually demand far chunkier hardware. Now, the headline number that made everyone
01:26raise an eyebrow is the cost. Benchmarks show medium 3 reaches more than 90% of Claude Sonnet's overall
01:33benchmark score. Yet Mistral quotes just $0.40 per million input tokens and $20.80 per million output tokens
01:41when you hit their API. For comparison, Sonnet is listed at $3 per million input and $15 per million
01:47output. Some of Mistral's own research material even shows an alternative rate card of $2 per million
01:54output. So the exact figure depends on which skew or deployment path you pick, but either way,
01:59you're staring at something in the neighborhood of an 8x price cut. That's wild in a market where model
02:05bills blow up faster than GPU stock on launch day. Performance claims always need receipts and Mistral
02:12came armed. On Human Eval and MultiPLE, the two coding benchmarks everyone loves to quote medium 3
02:18matches or beats Claude Sonnet and GPT-4.0. Third-party human evaluations from Surge show it winning 82%
02:26of coding scenarios against Llama 4 Maverick and nearly 70% against Cohere Command A. It's not just
02:33about code either. Pause multilingual tasks at it and you get higher win rates over Llama 4 Maverick
02:40in English 67%, French 71, Spanish 73, and Arabic 65. On multimodal reasoning, numbers like 0.953 on
02:53Dock VQA, 0.937 on AI2D, and 0.826 on Chart QA, pop up impressive because multimodal is still where
03:04many mid-sized models all that horsepower turns out to be especially handy for STEM workloads. Medium 3
03:12doesn't lock up while chewing on giant math proofs or engineering docs, and it compiles code fast enough
03:17that dev teams in finance, energy, and healthcare have already plugged data versions into production
03:22pipelines. A couple of those early testers are reportedly letting the model pre-train continuously
03:27on proprietary data, then fine-tuning in quick bursts when requirements shift, effectively running
03:33an in-house feedback loop without the headache of starting from scratch every time. That adapt-as-you-go
03:39angle is part of Mistral's pitch. Don't pick between black box SaaS fine-tuning or a DIY deployment,
03:46just blend both. Money matters too. And Medium 3 isn't only cheap compared to Anthropics lineup,
03:53it also beats DeepSeq V3, which until now enjoyed the reputation of being the cost-efficiency champ.
04:00That allows small teams to kick the tires through the API, then graduate to a self-hosted image when
04:06the CFO starts breathing down their neck about data residency or vendor lock-in. Astral even calls Medium
04:123 a proprietary model, so no MIT-style license, but they've kept everything flexible, throw it in
04:18Mistral LaPlatefmum, light it up on Amazon SageMaker, or wait a few weeks for IBM Watson X, NVIDIA NIM,
04:26Azure AI Foundry, and Google Cloud Vertex integrations to go live. Whichever route you choose,
04:31the company insists, you'll still be able to slide their weights onto your own GPU stack if you want
04:37total control. The under-the-hood strategy is all about hybrid deployments. You can keep inference
04:44in a private subnet, slam a low-latency tenant in a public region for burst traffic, or fork the entire
04:51thing and run it fully on-prem. Because we're talking about a French firm operating under GDPR
04:58and soon the EU AI Act, data governance boxes get ticked pretty aggressively. Audit logs, fine-grained
05:06ACLs, memory-based personalization, and the ability to unplug from the cloud completely all come baked
05:13into the architecture. That's gold for banks, hospitals, and utilities who live and die by
05:19regulations. And that credibility boost lands at the perfect moment because Medium 3 is already the
05:25motor under Le Chat Enterprise, the customer-facing layer that Mistral hopes will move it from cool
05:31research shop to everyday fixture inside big company workflows. Social Media Platform delivered
05:38exactly what employees wanted, with driving underpass to laser-cut your immersive public speaking
05:43programs down to minutes you don't have to, but bolts on everything a CIO circles in red when they read
05:49an AI RFP. When you wire it up to Google Drive, Gmail, Calendar, Microsoft SharePoint, OneDrive, or whatever
05:55connector they ship next, it does a single-pass search across all those silos, then snapshots its sources
06:02so compliance knows exactly where each sentence came from. If the file is a 60-page PDF, auto-summary skims
06:08the cruft, and the model drops links you can audit. The same Medium 3 stack powers a no-code agent builder,
06:14drag a few blocks together, and suddenly the assistant can pull a contract, update the CRM,
06:19and ping legal without anyone writing a cron job. Because everything sits on Medium 3's cheaper token
06:25pricing, finance teams finally get a clean line item instead of three overlapping SKUs. Security is the
06:32non-negotiable. Lachette Enterprise will run a SaaS in Mishra's own cloud, but you can flip the switch
06:38to a single-tenant region, a private VPC, or an on-prem rack, and keep the data inside your firewall.
06:45Access controls are inherited from the source apps, so a board deck locked to the CFO stays locked.
06:51Full audit logs ship out for SOC 2 and ISO paperwork, which matters if you're, say, a French bank living
06:57under GDPR and the incoming EU AI app. That EU angle quietly gives Mistral an edge with customers,
07:05who are wary of routing sensitive traffic through U.S. clouds or Chinese open-weight models.
07:11Medium 3 and Le Chat Enterprise grabbed the headlines, but Mistral's catalog is getting crowded.
07:16There's Mistral-Large 2, its GPT-4 class flagship. Pictstral-Large for images and docs.
07:24Codeistral for pure code generation, the Liz-Ministral Edge models that squeeze onto phones,
07:31and the Arabic-focused Mistral Saba. Alright. In March, they even shipped Mistral OCR,
07:38an API that turns any PDF into plain text so Medium 3 can actually read the stuff, legal still prints.
07:46Some models are wide open under Apache 2.0, the newest higher-end weights, including Medium 3,
07:52stay proprietary so Mistral can lock down licensed content and offer paid SLAs. That two-track approach
07:59is how they square their original openness slogan with the realities of enterprise contracts.
08:04If the roadmap feels impatient, look at the cap table. Since June 2023, the company has raised about
08:111 billion euros, including a $112 million seed that was Europe's largest on record, a $415 million Series A,
08:19led by Andreessen Horowitz and a $600 million mix of equity and debt last summer that parked the valuation
08:26at roughly $6 billion. Microsoft chipped in 15 million euros and hosts the weights on Azure,
08:33while NVIDIA, Cisco, Samsung, and IBM all took smaller slices. On the revenue side, paid API usage and
08:41the 14.99 per month Le Chat Pro plan are growing, but insiders still peg annual sales in the low
08:48eight digits. So scaling fast is existential. Partnerships help. The French army, press agency AFP,
08:56letting Le Chat query every story since 1983. Shipping giant CMA, CGM, and defense startup Helsing,
09:05all signed up. Even President Macron pitched Le Chat on TV last week, telling viewers to download the
09:12French app instead of importing ChatGPT. That kind of home-field backing doesn't guarantee market share,
09:19but it keeps the spotlight bright, while Mistral chases the numbers needed for the IPO that CEO
09:24Arthur Mensch keeps hinting at. All that context explains why Medium 3 is more than a mid-sized
09:31curiosity. It hits the sweet spot between small enough to run on four GPUs and smart enough to
09:37finish real work, and it does it at roughly one-eighth the cost of Anthropik's Claude Sonnet for the same
09:43token count. For dev teams watching cloud bills spike, a 60% benchmark tie with GPT-4 class models for
09:50pennies on the dollar is a conversation starter. For risk officers, the EU jurisdiction and on-prem
09:56option tick political and regulatory boxes OpenAI can't check yet. Looking ahead, the company is
10:03openly teasing a large release. If Medium is already closing the gap with openweight flagships like
10:10Llama 4 Maverick, a true Mistral Large 3 could yank the high-end leaderboard again, but Mistral's
10:16bigger challenge is commercial, not technical. It has to turn brand buzz and government endorsements
10:22into sustainable post-GAP revenue before acquisition rumors start looking more attractive than the
10:29NASDAQ bell. Mensch told reporters at Davos that the startup is not for sale and a public listing is
10:36the plan, but those words only hold if revenue catches up to that $6 billion price tag.
10:42For now, Medium 3 plus LeChat Enterprise give them a real shot. If you're tracking inference cost,
10:47on-prem compliance, or just need an LLM that speaks French and Arabic as well as English,
10:52keep an eye on the Google Cloud Marketplace listing that went live yesterday with Azure AI and AWS
10:58Bedrock slots coming next. And if you've already tried the Publicly Chat web app, remember,
11:03the model behind the curtain now writes code, summarizes PDFs, and cross-references your SharePoint
11:09without launching a dozen plugins or draining your GPU budget. Whether that's enough to vault Mistral
11:15into the same usage tier as OpenAI is the billionaire question, but at least now they're swinging
11:21with heavyweight gloves. Thanks for watching, catch you in the next one.

Recommended